Application of electrochemical impedance spectroscopy for prediction of fuel cell degradation by LSTM neural networks

Due to the new constraints on clean energy production, fuel cell (FC) devices are taking a predominant role in sustainable mobility due to their high efficiency and zero-emission - if hydrogen is used as fuel - compared to the internal combustion engine. They are already used in a multitude of applications (e.g. hybrid electric vehicles, portable devices, marine powerplants and stationary off-grid or UPS) but proper use of this technology is required to avoid fast degradation and reliability issues. Reliability is a fundamental aspect of fuel cell applications, making this technology the most suitable choice between internal combustion engines or battery-powered vehicles. The target of this paper is to find a suitable fault prediction algorithm using a Long Short-Term Memory (LSTM) neural network able to measure the degradation of a FC stack using a few and fast point-to-point measurements by the Electrochemical Impedance Spectroscopy (EIS). The final goal is to predict the degradation curve, i.e. the health status, represented by the cell voltage through cycles of usage, of a fuel cell.